Offline Speech to Text 0.14.0
Detailed troubleshooting now offers an explicit choice of one hour or 24 hours. Both expire automatically. The local detailed log keeps up to 1 MiB of recent metadata; basic logging keeps its 64 KiB limit. Heavy use can replace older records before the selected interval ends.
Diagnostic timestamps now record when an event occurs, and result logs distinguish decoded text from Android callback delivery. Callback return does not prove that the receiving app inserted the text.
The microphone meter reuses one animation and stops it when its view detaches. This removes repeated animation allocations. It does not establish faster inference or lower energy consumption. Speech recognition, models, thread defaults, and stored settings are preserved.
Diagnostics stay off by default. Audio, transcripts, editor contents and dictionary terms are excluded, and nothing is uploaded automatically.
Comments
No comments yet. Start the conversation.